Very high resolution remote sensing image classification with SEEDS-CNN and scale effect analysis for superpixel CNN classification

Lv, Xianwei; Ming, Dongping; Chen, Yangyang; Wang, Min

doi:10.1080/01431161.2018.1513666

Cited by 108 publications

(64 citation statements)

References 43 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…More recently, algorithms employing deep learning (DL) (i.e., Deep Neural Networks (DNNs)) [12] have also become very popular for LULC classification [13]. Recently, several powerful architectures of DL models have been developed for the classification of RS images [14][15][16]. It is, however, still arguable how well these DL algorithms perform against ensemble algorithms and SVM for the purpose of urban LULC classification.…”

Section: Machine Learning Classifiers For Object-based Classificationmentioning

confidence: 99%

Comparing Deep Neural Networks, Ensemble Classifiers, and Support Vector Machine Algorithms for Object-Based Urban Land Use/Land Cover Classification

2019

View full text Add to dashboard Cite

With the advent of high-spatial resolution (HSR) satellite imagery, urban land use/land cover (LULC) mapping has become one of the most popular applications in remote sensing. Due to the importance of context information (e.g., size/shape/texture) for classifying urban LULC features, Geographic Object-Based Image Analysis (GEOBIA) techniques are commonly employed for mapping urban areas. Regardless of adopting a pixel- or object-based framework, the selection of a suitable classifier is of critical importance for urban mapping. The popularity of deep learning (DL) (or deep neural networks (DNNs)) for image classification has recently skyrocketed, but it is still arguable if, or to what extent, DL methods can outperform other state-of-the art ensemble and/or Support Vector Machines (SVM) algorithms in the context of urban LULC classification using GEOBIA. In this study, we carried out an experimental comparison among different architectures of DNNs (i.e., regular deep multilayer perceptron (MLP), regular autoencoder (RAE), sparse, autoencoder (SAE), variational autoencoder (AE), convolutional neural networks (CNN)), common ensemble algorithms (Random Forests (RF), Bagging Trees (BT), Gradient Boosting Trees (GB), and Extreme Gradient Boosting (XGB)), and SVM to investigate their potential for urban mapping using a GEOBIA approach. We tested the classifiers on two RS images (with spatial resolutions of 30 cm and 50 cm). Based on our experiments, we drew three main conclusions: First, we found that the MLP model was the most accurate classifier. Second, unsupervised pretraining with the use of autoencoders led to no improvement in the classification result. In addition, the small difference in the classification accuracies of MLP from those of other models like SVM, GB, and XGB classifiers demonstrated that other state-of-the-art machine learning classifiers are still versatile enough to handle mapping of complex landscapes. Finally, the experiments showed that the integration of CNN and GEOBIA could not lead to more accurate results than the other classifiers applied.

show abstract

Section: Machine Learning Classifiers For Object-based Classificationmentioning

confidence: 99%

Comparing Deep Neural Networks, Ensemble Classifiers, and Support Vector Machine Algorithms for Object-Based Urban Land Use/Land Cover Classification

2019

View full text Add to dashboard Cite

show abstract

“…The superpixel segmentation results are propitious for center point voters to reduce the deep features that may lead to misclassifications in the process of feature extraction. In the paper [44], SEEDS has proven that it is the most suitable superpixel segmentation for CNNs, better than other commonly used superpixel segmentation methods. Therefore, for SEEDS-CNN, one voter would be sufficient.…”

Section: Effectiveness Of Rmv-cnn For Vhri Classificationmentioning

confidence: 99%

A New Method for Region-Based Majority Voting CNNs for Very High Resolution Image Classification

Ming

et al. 2018

Remote Sensing

Self Cite

View full text Add to dashboard Cite

Conventional geographic object-based image analysis (GEOBIA) land cover classification methods by using very high resolution images are hardly applicable due to their complex ground truth and manually selected features, while convolutional neural networks (CNNs) with many hidden layers provide the possibility of extracting deep features from very high resolution images. Compared with pixel-based CNNs, superpixel-based CNN classification, carrying on the idea of GEOBIA, is more efficient. However, superpixel-based CNNs are still problematic in terms of their processing units and accuracies. Firstly, the limitations of salt and pepper errors and low boundary adherence caused by superpixel segmentation still exist; secondly, this method uses the central point of the superpixel as the classification benchmark in identifying the category of the superpixel, which does not allow classification accuracy to be ensured. To solve such problems, this paper proposes a region-based majority voting CNN which combines the idea of GEOBIA and the deep learning technique. Firstly, training data was manually labeled and trained; secondly, images were segmented under multiresolution and the segmented regions were taken as basic processing units; then, point voters were generated within each segmented region and the perceptive fields of points voters were put into the multi-scale CNN to determine their categories. Eventually, the final category of each region was determined in the region majority voting system. The experiments and analyses indicate the following: 1. region-based majority voting CNNs can fully utilize their exclusive nature to extract abstract deep features from images; 2. compared with the pixel-based CNN and superpixel-based CNN, the region-based majority voting CNN is not only efficient but also capable of keeping better segmentation accuracy and boundary fit; 3. to a certain extent, region-based majority voting CNNs reduce the impact of the scale effect upon large objects; and 4. multi-scales containing small scales are more applicable for very high resolution image classification than the single scale.

show abstract

“…As convolutional neural networks show strong advantages in the field of natural images, more and more researchers have tried to apply convolutional neural networks to the field of remote sensing images, with some progress made in the segmentation and recognition of remote sensing images. Lv et al classified remote sensing images with SEEDS-CNN and scale effectiveness analysis [20]. Chen et al applied multi-scale CNN and scale parameter estimation in land cover classification [21].…”

Section: Introductionmentioning

confidence: 99%

BRRNet: A Fully Convolutional Neural Network for Automatic Building Extraction From High-Resolution Remote Sensing Images

et al. 2020

View full text Add to dashboard Cite

Building extraction from high-resolution remote sensing images is of great significance in urban planning, population statistics, and economic forecast. However, automatic building extraction from high-resolution remote sensing images remains challenging. On the one hand, the extraction results of buildings are partially missing and incomplete due to the variation of hue and texture within a building, especially when the building size is large. On the other hand, the building footprint extraction of buildings with complex shapes is often inaccurate. To this end, we propose a new deep learning network, termed Building Residual Refine Network (BRRNet), for accurate and complete building extraction. BRRNet consists of such two parts as the prediction module and the residual refinement module. The prediction module based on an encoder–decoder structure introduces atrous convolution of different dilation rates to extract more global features, by gradually increasing the receptive field during feature extraction. When the prediction module outputs the preliminary building extraction results of the input image, the residual refinement module takes the output of the prediction module as an input. It further refines the residual between the result of the prediction module and the real result, thus improving the accuracy of building extraction. In addition, we use Dice loss as the loss function during training, which effectively alleviates the problem of data imbalance and further improves the accuracy of building extraction. The experimental results on Massachusetts Building Dataset show that our method outperforms other five state-of-the-art methods in terms of the integrity of buildings and the accuracy of complex building footprints.

show abstract

Very high resolution remote sensing image classification with SEEDS-CNN and scale effect analysis for superpixel CNN classification

Cited by 108 publications

References 43 publications

Comparing Deep Neural Networks, Ensemble Classifiers, and Support Vector Machine Algorithms for Object-Based Urban Land Use/Land Cover Classification

Comparing Deep Neural Networks, Ensemble Classifiers, and Support Vector Machine Algorithms for Object-Based Urban Land Use/Land Cover Classification

A New Method for Region-Based Majority Voting CNNs for Very High Resolution Image Classification

BRRNet: A Fully Convolutional Neural Network for Automatic Building Extraction From High-Resolution Remote Sensing Images

Contact Info

Product

Resources

About